|
(11) | EP 0 595 353 A1 |
| (12) | EUROPEAN PATENT APPLICATION |
|
|
|
|
|||||||||||||||||||||||||||
| (54) | A method of and an apparatus for generating a normalizing curve |
| (57) A normalizing curve (Crx) is generated by correcting a reference normalizing curve. The character of the normalized
curve depends on dummy shadow and highlight points (A₄, A₃) as well as shadow and
highlight points (A₂, A₁), which are determined as a function of the statistical distribution
of densities on an original image. Color original image signals representative of
the original image are converted through the normalizing curve to obtain converted
color image signals in which a gray balance in the intermediate density region is
corrected. |
BACKGROUND OF THE INVENTION
Field of the Invention
[0001] The present invention relates to a method of and an apparatus for generating a normalizing curve which is used in an image processing apparatus which processes an image such as a process scanner.
Description of the Prior Art
[0002] In a process color scanner and the like, after an image of a color original is subjected to color separation to thereby obtain density signals for the respective color components, the density signals are converted into halftone dot signals with which recording of a reproduced image is performed. If there is a hue deviation in the original, color separation conditions are changed at the discretion of an operator so that the gray balance will be maintained in the reproduced image.
[0003] In a conventional method of automatically altering color separation conditions as that disclosed in U.S. Patent No. 4,984,071 filed by the inventor of the present invention, a reference density point is determined for each color component based on density values of the brightest portion and the darkest portion of the original (i.e., highlight and shadow density values), a gradation conversion curve (i.e., normalizing curve) is established so that these density values have predetermined signal levels, and different density distributions of image data are normalized for the respective images. This makes it possible that peaks in the distributions of the density values of a bright portion and a dark portion are corrected and an image entirely having a tone of a certain color (hue deviation) is corrected.
[0004] However, in such a correction method as above (hereinafter "first conventional method"), if there is a hue deviation only in a shadow portion due to a fact that the maximum value which can be expressed by a color film, i.e., the original, is different among the respective color components, it is difficult to perform appropriate correction. That is, since the first conventional method uniformly performs correction over a wide density range, if correction to remove the hue deviation in the shadow portion is performed, a low and intermediate density region will be also corrected and the gray balance in the low and intermediate density region will be lost.
[0005] To deal with this, the inventor of the present invention has proposed a normalizing curve generation method as that disclosed in U.S. Patent No. 5,117,293 (hereinafter "second conventional method"). In the second conventional method, in addition to the highlight and the shadow density values mentioned above, a dummy shadow density value is determined in a predetermined manner so that a normalizing curve is generated which changes along a straight line which passes through a first and a second points which are determined by said highlight density value and said shadow density value on a coordinates plane and through a third point which is determined by said first point and said dummy shadow point in the low and intermediate density region.
[0006] According to the second conventional method, by properly setting the third point (i.e., dummy shadow density value), the hue deviation near the shadow density value is corrected and a normalizing curve is changed without affected by the amount of correction made in the shadow density range in the low and intermediate density region.
[0007] Still, due to characteristics of a color film, i.e., an original, if there are differences between the respective color components in terms of a ratio of change in a density with respect to a logarithmic value of the quantity of incident light in a section from the low density range to the intermediate density range, it is impossible to appropriately correct the gray balance in the intermediate density region by the second conventional method, not to mention by the first conventional method.
[0008] In other words, since correction of the hue deviation over all density regions is performed in the first conventional method while correction only in the vicinity of the shadow density value is performed in the second conventional method, it is impossible in both the first and the second conventional methods to perform appropriate correction of the gray balance only in the intermediate density region, or in the intermediate density region and in the vicinity of the shadow density value, or in the intermediate density region and in the vicinity of the highlight density value.
[0009] An object of the present invention is to solve such a problem and to offer a normalizing curve generation method which makes it possible to perform correction of the gray balance in an intermediate density region which has been heretofore impossible.
SUMMARY OF THE INVENTION
[0010] The present invention is directed to a method of converting original color image signals Dx into converted color image signals DNx suitable for image processing in a color image processor. The original color image signals Dx represent optical densities of an original color image for each pixel and for respective color components.
[0011] According to the present invention, the method comprises the steps of: (a) specifying gray pixels on the original color image; (b) averaging the original color image signals Dx for respective color components in each gray pixel to obtain an average signal DM for each gray pixel; (c) dividing a range of the average signal DM into a plurality of classes DMi; (d) classifying the average signal DM representative of the gray pixels into the plurality of classes DMi; (e) accumulating the original color signals Dx for each color component in each class DMi of the average signal DM to obtain accumulated density signal ΣDx for each color component; (f) generating a sum signal Σ(ΣDx)M representative of a sum of the accumulated density signal ΣDx for each color component within a predetermined intermediate density section DMk ≦ DM ≦ DMh; (g) counting number ΣNpn of pixels whose original color signals Dx are included in the intermediate density section DMk ≦ DM ≦ DMh; (h) dividing the sum signal Σ(ΣDx)M for each color component by the number ΣNpn of pixels to obtain accumulated average signals Mdx for respective color components; (i) averaging the accumulated average signals Mdx for respective color components to obtain a gray average signal Mda; (j) generating difference signals
between the accumulated average signals Mdx and the gray average signal Mda; (k) modifying a predetermined reference signal-conversion characteristic C as a function of the difference signals
to obtain modified signal-conversion characteristics Crx for respective color components; and (1) converting the original color image signals Dx by the modified signal-conversion characteristics Crx to obtain the converted color image signals DNx for respective color components.
[0012] The difference signals
well represent relative deviations of respective color components in the intermediate density section, and therefore, the gray balance in the intermediate density section is attained by using the modified signal-conversion characteristics Crx which are obtained as a function of the difference signals
.
[0013] In an aspect of the present invention, signal-conversion characteristic for converting original color image signals Dx are obtained by the steps of: (a) determining a highlight density DHx and a shadow density DSx in a possible density range of the original color image signals Dx; (b) determining dummy highlight densities CDHx and dummy shadow densities CDSx for respective color components in the possible density range of the original color image signals Dx; and (c) determining signal-conversion characteristics Crx for respective color components representative of signal-conversion curves defined on a coordinate plane satisfying the conditions that; I) the signal-conversion curves pass through; first points (DHx, px) defined by the highlight densities DHx and first predetermined signal values px, and second points (DSx, qx) defined by the shadow densities DSx and second predetermined signal values qx, and II) part of the signal-conversion curves defined for a predetermined intermediate density section coincide with or approximate to lines connecting; third points (CDHx, px) defined by the dummy highlight densities CDHx and the first predetermined signal values px, and fourth points (CDsx, qx) defined by the dummy shadow densities CSHx and the second predetermined signal values qx.
[0014] The first and second points are effective for determining the feature of the signal-conversion curves in highlight and shadow sections, while the third and fourth points are effective for determining the feature of the signal-conversion curves in the intermediate density section.
[0015] Accordingly, an object of the present invention is to ensure the gray balance in an intermediate density region on a color image.
[0016] These and other objects, features, aspects and advantages of the present invention will become more apparent from the following detailed description of the present invention when taken in conjunction with the accompanying drawings.
BRIEF DESCRIPTION OF THE DRAWINGS
[0017]
Fig. 1 is a block diagram showing an overall structure of a process color scanner to which an embodiment of the present invention is applicable;
Fig. 2 is a block diagram showing the structure of a normalizing curve generation part of the embodiment of Fig. 1;
Fig. 3 is a flow chart showing operations of a reference normalizing curve setting part of the normalizing curve generation part of Fig. 2;
Fig. 4 is a flow chart for explaining procedures of data processing at a step S100 of Fig. 3;
Fig. 5 is a flow chart showing operations of an intermediate density region correction part of the normalizing curve generation part of Fig. 2;
Fig. 6 is a view showing an example of an average density frequency histogram;
Fig. 7 is a view showing an example of a cumulative density histogram;
Fig. 8 is a view showing an example of a cumulative relative frequency histogram;
Fig. 9 is an explanatory diagram for finding average density values in a highlight side density section and a shadow side density section;
Fig. 10 is a graph showing an example of a reference normalizing curve;
Fig. 11 is a view showing an example of a cumulative relative frequency histogram about gray region pixels;
Fig. 12 is an explanatory diagram for finding an average density value in a specified density section of the intermediate density region;
Fig. 13 is a view for explaining the nature of a correction normalizing curve;
Fig. 14 is a graph showing an example of the correction normalizing curve; and
Fig. 15 is a graph showing a relation between a coefficient t and a density value D.
DESCRIPTION OF THE PREFERRED EMBODIMENTS
〈A. Overall Structure and General Operation of Embodiment〉
[0018] Fig. 1 is a block diagram showing an overall structure of a process color scanner 100 to which an embodiment of the present invention is applicable.
[0019] A color original film 1 is wound around an original drum 2. In a faced relation to the original drum 2, a pick-up head 3 is disposed. By translation of the original drum 2 in a direction α (i.e., main scanning direction) and translation of the pick-up head 3 in a direction β (i.e., subscanning direction), the original film is scanned in both the main scanning direction and the subscanning direction, whereby the image of the original film 1 is photoelectrically read pixel by pixel in the order of scanning lines by the pick-up head 3.
[0020] The image information about the original film 1 is logarithmically converted by a logarithm convertor 4 for each one of the color components blue (B), green (G) and red (R) and then converted by an A/D conversion circuit 5 into digital density signals DB, DG and DR for the respective color components B, G and R.
[0021] The respective density signals Dx thus created in correspondence with the respective pixels (x=B, G, R; In the description hereinafter, the symbol "x" is used to indicate x=B, G, R. The symbol "x" will not be therefore described again) are supplied to a normalizing conversion part 200.
[0022] The normalizing conversion part 200 is comprised of a dividing circuit 201, a normalizing curve generation part 210 and a look up table (LUT) 202. During prescanning, the dividing circuit 201 is connected to the normalizing curve generation part 210 side as well, and the normalizing curve generation part 210 generates a reference normalizing curve by a method which will be described later based on the inputted density values (i.e., density signals Dx) and appearance frequencies of the respective pixels. Information about the reference normalizing curve is loaded in the LUT 202 for each color component as numeric data.
[0023] During scanning, the dividing circuit 201 is disconnected from the normalizing curve generation part 210 side and connected to the LUT 202 side. The density signals Dx of the respective pixels that were read are normalized based on the data which are stored in the LUT 202 so as to be changed to normalized density signals DNx which will be given to a next stage element, i.e., a first color computation part 6.
[0024] The first color computation part 6 performs color computation based on the normalized density signals DNx to thereby generate color separation signals Y, M, C and K. These color separation signals are converted into halftone dot signals Y, M, C and K in a halftone dot generation part 7. The halftone dot signals are combined in a time sharing manner by a selector 8 and provided to an output circuit 9.
[0025] A recording head 10 is equipped with a laser light source so as to ON/OFF modulate a laser beam in accordance with a modulation signal which is received from the output circuit 9. A photosensitive film 12 is wound around a recording drum 11. By combination of translation of the drum 11 in the direction α (i.e., main scanning direction) and translation of the recording head 10 in the direction β (i.e., subscanning direction), the photosensitive film 12 is exposed by the laser beam pixel by pixel in the order of the scanning lines.
[0026] On the other hand, independently of prescanning and scanning, the color separation signals Y, M, C and K outputted from the first color computation part 6 are stored for each pixel in a monitoring frame memory 13 and converted again into color signals B, G and R in a second color computation part 14 before displayed at a color monitor 15. Thus, the operator can manipulate image processing performed at the normalizing conversion part 200 and the first color computation part 6 while monitoring the processing.
[0027] An information processing apparatus 16 comprises a CPU 17 and a memory 18 for storing a system program and controls operations of the normalizing conversion part 200, the first color computation part 6, the monitoring frame memory 13 and the second color computation part 14. A console 19 is provided to manually input various commands and data into the information processing apparatus 16.
[0028] In the following, description will be given on a case where the color separation images Y, M, C and K are recorded positive on the photosensitive film 12.
〈B. Structure and Operation of Normalizing Curve Generation Part〉
[0029] Fig. 2 is a block diagram showing an embodiment of the structure of the normalizing curve generation part 210 of the normalizing conversion part 200 (Fig. 1). The normalizing curve generation part 210 is essentially comprised of an input part 220, a reference normalizing curve setting part 230 and an intermediate density region correction part 240. The detailed structure of these parts will be described later with reference to Figs. 3, 4 and 5 each showing a flow chart explaining generation of a normalizing curve.
[0030] In the present embodiment, a reference normalizing curve which is created by the second conventional method described in U.S. Patent No. 5,117,293 is used. The reference normalizing curve is determined as a function of highlight and shadow density values and a dummy shadow density value for appropriately correcting the hue deviation in the vicinity of the shadow density value. In low and intermediate density regions, the reference normalizing curve is free from influence of the amount of the hue deviation in the shadow density range. In the present embodiment, the reference normalizing curve is corrected appropriately to obtain a new normalizing curve in order to correct the gray balance in the intermediate density region which may be lost due to characteristics of a color original film. The details of the gray balance correction will be described later.
[B-1. Generation Method of Reference Normalizing Curve]
[0031] Figs. 3 and 4 are flow charts showing how the normalizing curve generation part 210 of Fig. 2 establishes a reference normalizing curve. In the following, successive stages of establishing process of a reference normalizing curve will be described with reference to these flow charts.
[B-1-1. Calculation of Average Density Value]
[0032] The image signals which are created by prescanning of the original film 1 are converted into the digital density signals Dx by the logarithm convertor 4 and the A/D conversion circuit 5 (Fig. 1) as described earlier. Following this, through the dividing circuit 201, the density signals Dx of the respective pixels are stored for each one of the color components B, G and R in the frame memory 221 (Fig. 2) of the input part 220 which is disposed in the normalizing curve generation part 210 (Step S1 of Fig. 3).
[0033] If the original film 1 is a large size film, the density signals Dx are stored in the frame memory 221 while subtracting the pixels.
[0034] During reading of the density signals Dx which are stored in the frame memory 221 by a next stage element, i.e., the reference normalizing curve setting part 230 or the intermediate density region correction part 240, a coordinates value of a predetermined image portion is generated by a coordinates generation part 222, whereby an image within a predetermined range is designated as an image-to-be-processed and the density signals Dx of corresponding pixels are sent to the next stage element for each pixel.
[0035] A next step S100 is a sub routine for finding a highlight side average density value DHAx and a shadow side average density value DSAx from these density signals Dx. Fig. 4 is a flow chart showing the details of the step S100.
[0036] At a step S101, based on the density signals Dx, for each pixel given from the input part 220, an average density value
is calculated by a first histogram processing part 231 (Fig. 2). This calculation is performed on all pixels which are sent from the input part 220. An average density frequency histogram generation part 231N generates an average density frequency histogram of each color component in which the range of the average density value DM is measured along a horizontal axis and the number of pixels NP is measured along a vertical axis. Fig. 6 shows an example of an average density frequency histogram. In Fig. 6, a class mean value (i.e., class value) is shown as DMi (i=1 to n).
[0037] At a step S102, cumulative density histogram generation parts 231B, 231G and 231R which are disposed in correspondence with the respective color components accumulates the density signals Dx of the respective pixels in each class of the respective average density frequency histograms of the respective color components B, G and R. After this computation is performed, cumulative density histograms for the respective color components are generated in each of which the class value DMi (i= 1 to n) is measured along a horizontal axis and a cumulative density value ΣDx which corresponds to the pixels which are included in each class is measured along a vertical axis. Fig. 7 shows an example of the cumulative density histograms. For each color component x=B, G, R, such a cumulative density histogram is generated.
[0039] Next, at a step S103, based on the average density frequency histogram shown in Fig. 6, a first cumulative processing part 232 generates a cumulative relative frequency histogram as that shown in Fig. 8 in which the class value DMi (i=1 to n) is measured along a horizontal axis and a relative frequency RN (%) of the pixels accumulated from a lower density side is measured along a vertical axis. The cumulative relative frequency histogram varies from 0% to 100% in the range between the minimum generated density DMmin and the maximum generated density DMmax. It is to be noted that the cumulative relative frequency histogram of Fig. 8 approximates a curve on a premise that the width of each class is sufficiently narrow.
[0040] In a first section setting part 233 shown in Fig. 2, predetermined cumulative density appearance rates RNh and RNS are registered. Based on these cumulative density appearance rates RNh and RNS, from the cumulative relative frequency histogram of Fig. 8, temporary highlight average density value DMH and a temporary shadow average density value DMS are respectively calculated (Step S104). Based on the temporary highlight average density value DMH and the temporary shadow average density value DMS, information about a highlight side density section and a shadow side density section whose average density values are to be calculated is created and given to a first average density computation part 234.
[0041] The cumulative density appearance rates RNh and RNS are found in advance by analysis of a number of sample originals as values from which an optimum highlight point and an optimum shadow point are statistically determined. For instance, the cumulative density appearance rates RNh and RNS are around 1% and 98%, respectively.
[0042] The first average density computation part 234 calculates a highlight side average density value DHAx and a shadow side average density value DSAx for the respective color components of a density section which is designated by the first section setting part 233 (Step S105).
[0043] That is, in the cumulative density histogram of each color component shown in Fig. 7, as indicated at slashed portions, with respect to the highlight side, the cumulative density values ΣDx of the pixels included in the density section equal to or smaller than the temporary highlight average density value DMH (DMmin ≦ DM ≦ DMH) are averaged. With respect to the shadow side, the cumulative density values ΣDx of the pixels included in the density section equal to or larger than the temporary shadow average density value DMS ((DMS ≦ DM ≦ DMmax) are averaged.
[0044] Assume that the total number of the pixels which belong to the highlight side density section above is ΣNPH, the total of the cumulative density values within the same highlight side density section is Σ(ΣDx)H, the total number of the pixels which belong to the shadow side density section above for the respective color components is ΣNPS and total of the cumulative density values within the same shadow side density section for the respective color components is Σ(ΣDX)S, the highlight side average density value DHAx and the shadow side average density value DSAx are defined by Eq. 1 and Eq. 2 below, respectively.
For example, when x is B in Fig. 9 and the temporary highlight density value DMH is DM5, from Eq. 1,
where NPi (i=1 to 5) is the number of the pixels whose average density values DM belong to the class DMi. The value NPi is easily found from the average density frequency histogram of Fig. 6.
[0045] This completes the sub routine of the step S100. Following this, the sequence returns to the main routine shown in Fig. 3.
[B-1-2. Determination of Highlight Reference Value and Shadow Reference Value]
[0047] Based on the highlight side average density value DHAx and the shadow side average density value DSAx above, a reference density value which is commonly applied to the respective color components (i.e., a highlight reference density value DHO and a shadow reference density value DSO) is calculated by a reference density value setting part 235 (Step S2).
[0048] Where a standard highlight density value and a standard shadow density value which are preliminarily determined are DHF and DSF, respectively, the highlight reference density value DHO is obtained as a predetermined weighted mean value of the minimum value of the highlight side average density value DHAx and the standard highlight density value DHF, whereas the shadow reference density value DSO is obtained as a predetermined weighted mean value of the maximum value of the shadow side average density value DSAx and the standard shadow density value DSF.
[0049] Alternatively, the highlight reference density value DHO and the shadow reference density value DSO may be replaced by an average value of the highlight side average density values DHAx of the respective color components and an average value of the shadow side average density values DSAx of the respective color components, respectively.
[B-1-3. Calculation of Gray Balance Correction Amount]
[0050] Next, a gray balance correction amount ΔDHx of a highlight portion and a gray balance correction amount ΔDSx of a shadow portion are calculated by the following Eqs. 3 and 4 (Step S3).
The values KH and KS are positive constants which are experimentally determined in advance, and the functions GH and GS are defined as, for example:
where
DHmax = MAX (DHAB, DHAG, DHAR)
DHmin = MIN (DHAB, DHAG, DHAR)
AH : preliminarily selected positive constance
m : preliminarily selected positive constance (for instance, "3")
As can be understood from Eqs. 3 to 6, the correction amount ΔDHx of the highlight portion is in proportion to (DHAx - DHO) while the correction amount ΔDSx of the shadow portion is in proportion to (DSAx - DSO). The value (DHAx - DHO) reflects the amount of the hue deviation in the highlight portion while the value (DSAx - DSO) reflects the amount of the hue deviation in the shadow portion.
[B-1-4. Determination of Highlight Density Value and Shadow Density Value]
[0051] At a next step S4, using the reference density values DHO and DSO and the correction amounts ΔDHx and ΔDSx, a highlight density value DHx and a shadow density value DSx for each color component are calculated by Eqs. 7 and 8 below.
[B-1-5. Determination of Dummy Shadow Density]
[0053] At a step S5 that follows, the operator judges the necessity of correction of the hue deviation only in the shadow portion of the original film 1.
[0054] If correction of the hue deviation only in the shadow portion is necessary, a dummy shadow density value setting part 236 sets dummy shadow density values DDSx for the respective color components at a common value using the shadow reference density value DSO (Step S6). That is,
.
[0055] On the other hand, if the original 1 is entirely faded, the dummy shadow density value setting part 236 determines the dummy shadow density values DDSx as the shadow density values DSx of the corresponding color components (Step S7). In short,
. At the step S7, the dummy shadow density value DDSx may not be set exactly equal to the shadow density value DSx. Instead, the dummy shadow density value DDSx may be set close to the shadow density value DSx.
[B-1-6. Determination of Reference Normalizing Curve]
[0056] A reference function setting part 237 specifies a reference normalizing curve for each color component using a predetermined nonlinear function Fx based on the highlight density value DHx, the shadow density value DSx and the dummy shadow density value DDSx (Step S8).
[0057] As an example of the function Fx which shows such a reference normalizing curve, nonlinear functions Fx as those defined by the following Eqs. 9 to 11 are determined beforehand, and information expressing these functions is stored in an inner memory of the reference function setting part 237.
where
p, q : constants set in the vicinity of the upper and the lower limits of a normalizing density range before color computation in the first color computation part 6
D : density value before normalization
DN : density value after normalization
u = DHx
v = DDSx
w = DSx
Fig. 10 is a view in which the configuration of a reference normalizing curve C which is specified by the nonlinear function Fx above is shown on a normalizing coordinates plane of density values. Fig. 10 concerns where the correction amount ΔDSx of the shadow portion has a positive value in Eq. 8. In the coordinates plane, a density value before normalizing Dx is measured along a horizontal axis while a density value after normalizing DN is measured along a vertical axis. The symbols px and qx respectively denote the values of the constants p and q with respect to a color component x.
[0058] As shown in Fig. 10, the function Fx (and hence the normalizing curve C which is generated by the same) has the following natures (i) to (iv):
(i) The reference normalizing curve C substantially passes through a point QH which is located at two-dimensional coordinates (DHx, px);
(ii) The reference normalizing curve C passes through a point QS which is located at two-dimensional coordinates (DSx, qx);
(iii) The reference normalizing curve C changes along a straight line L in a highlight
portion HLP but is off the straight line L in a shadow portion SDP. The reference
normalizing curve C deviates from the straight line L largely as the density value
Dx becomes large. The straight line L is a line from a point QD which is located at
two-dimensional coordinates (DDSx, qx) and the point QH which is mentioned above.
Since the amount of deviation in the shadow portion SDP is determined in accordance
with the amount of the hue deviation only in the shadow portion of the original film
1, using the reference normalizing curve of this embodiment shown as the curve C in
Fig. 10, it is possible to remove the hue deviation only in the shadow portion without
degrading the gray balance in the low and intermediate density region.
(iv) If a problem with the original film 1 is not just a hue deviation only in the shadow portion but fading of colors over the entire portion of the original film 1, the shadow density value DSx and the dummy shadow density value DDSx are set equal to each other. A normalizing curve in such a case is a straight line L0 which passes through the points QH and QS.
[0059] Normalizing characteristics expressed by the straight line L0 are capable of suppressing the density of a specific color component in both an intermediate density range MDP and the shadow portion SDP, maintaining the gray balance among the respective color components, and correcting the fading over the entire portion of the original film 1.
[0060] Hence, the nonlinear function Fx used in this embodiment is a function which is effective in correcting both a hue deviation only in the shadow portion and color fading over the entire portion of the original film 1. Thus, the nonlinear function Fx can be used for these various types of correction.
[B-2. Intermediate Density Region Correction Method]
[0061] Next, correction using the reference normalizing curve C performed at the intermediate density region correction part 240 of Fig. 2 will be described with reference to the flow chart of Fig. 5.
[0062] The reference normalizing curve C established at the reference normalizing curve setting part 230 in the manner above has excellent characteristics that it is possible to correct a hue deviation only in the shadow portion or in the entire film while ensuring the gray balance by appropriately setting the dummy shadow density value DDSx. However, due to an inherent nature of a color film, when a ratio of change of the density with respect to a logarithmic value of the amount of incident light in a section from the low density range to the intermediate density range is different between the color components, the gray balance in the intermediate density region cannot be properly corrected.
[0063] To correct the gray balance in the intermediate density region, in the manner described below, the nonlinear function Fx expressing the reference normalizing curve C is corrected.
[B-2-1. Detection of Gray Region Pixels]
[0064] The density signals Dx with respect to each pixel are inputted from the input part 220 to the intermediate density region correction part 240 (Step S11). A gray region pixel detection part 241 then judges whether the density signals Dx of each pixel are the density signals of a pixel which is included in a gray region of a color space (Step S12). If the pixels are found to belong to the gray region, the gray region pixel detection part 241 provides a second histogram processing part 242 with a signal indicating this (hereinafter "gray signal").
[0065] In a gray region pixel detection method, for example, as to the inputted density signals Dx of each pixel with respect to the respective color components, density differences Dd between the density values DB, DG and DR are calculated and whether the largest absolute value | Dd | of the differences is equal to or smaller than a predetermined threshold value is judged. If the largest absolute value | Dd | is equal to or smaller than the predetermined threshold value, the pixel is found to belong to the gray region of the color space. As the absolute values | Dd | of the density differences Dd are small, the density signals Dx with respect to G, B and R are closer to each other. Therefore, the pixel can be found to be a pixel in the gray region.
[B-2-2. Specifying Average Density Value in Medium Density Region]
[0066] Steps S13 to S17 are steps for finding an average density value of a gray region pixel in a predetermined density section of the intermediate density region. These steps approximately correspond to the process for finding the highlight side average density value DHAx and the shadow side average density value DSAx which have been described earlier in relation to the reference normalizing curve setting part 230 (flow chart of Fig. 4).
[0067] These steps are different from the process in that pixels to be subjected to density value averaging are not all pixels which are sent from the input part 220 but the pixels which are found by the gray region pixel detection part 241 to belong to the gray region (hereinafter "gray region pixels") and in that the density section in which the density values are to be averaged is a predetermined medium density region.
[0068] The density signals Dx from the input part 220 are also given to the second histogram processing part 242. The second histogram processing part 242, however, accepts the density signals Dx only when receives the gray signal from the gray region pixel detection part 241.
[0069] Based on the density signals Dx of the gray region pixels, the second histogram processing part 242 calculates an average density value
for each pixel. An average density frequency histogram generation part 242N generates a gray region average density frequency histogram for each color component in which a class indicating the range of the average density value DM is measured along a horizontal axis and the number of pixels NP is measured along a vertical axis (Step S13).
[0070] Cumulative density value table generation parts 242B, 242G and 242R accumulate the density values DB, DG and DR for the respective color components B, G and R of the pixels which are included in each class of the average density frequency histograms. After performing such computation, with a class value DMi (i=1 to n) measured along a horizontal axis and cumulative density values ΣDB, ΣDG and ΣDR corresponding to the pixels which are included in the respective classes measured along vertical axes, gray region cumulative density value tables for the respective color components are created (Step S14). In reality, the steps S13 and S14 are performed parallel to each other.
[0071] The gray region average density frequency histograms and the gray region cumulative density value tables correspond to the average density frequency histograms (Fig. 6) and the cumulative density histograms (Fig. 7), respectively, which have been described in relation to generation of the reference normalizing curve C. Since the gray region average density frequency histograms and the gray region cumulative density value tables are approximately the same as the average density frequency histograms (Fig. 6) and the cumulative density histograms (Fig. 7) except for the density range along the horizontal axis, the frequency in each class along the vertical axis, the value of the cumulative density and etc., the gray region average density frequency histograms and the gray region cumulative density value tables are not illustrated.
[0072] Next, based on the gray region average density frequency histograms, a second cumulative processing part 243 creates a gray region cumulative relative frequency histogram in which the class value DMi (i=1 to n) is measured along a horizontal axis and the relative frequency RN (%) of the pixels accumulated from the low density region is measured along a vertical axis (Step S15).
[0073] An example of the gray region cumulative relative frequency histogram is shown in Fig. 11. The configuration of the histogram changes from 0% to 100% within a range between a minimum generated density DMgmin and a maximum generated density DMgmax. On the premise that the class width is enough narrow likewise in Fig. 8, the gray region cumulative relative frequency histogram of Fig. 11 approximates a curve.
[0074] In a second section setting part 244, a predetermined range of a cumulative density appearance rate (RNk to RNh) in the intermediate density region is set. Average density values DMk and DMh are calculated which respectively correspond to the cumulative density appearance rates RNk and RNh of the gray region cumulative relative frequency histogram of Fig. 11. A density section of the intermediate density region defined by the average density values DMk and DMh is then determined, and the information regarding the density section is given to a second average density computation part 245 (Step S16).
[0075] The values of the cumulative density appearance rates RNk and RNh are determined by analysis of a number of sample originals so as to statistically find an optimum range. The cumulative density appearance rates RNk and RNh are appropriate values above and under 50%, e.g., about 40% and 60%.
[0076] The second average density computation part 245 divides the total of the cumulative density values which belong to the density section which is designated by the second section setting part 244 with the total of the appearance frequencies of the pixels which belong to the same density section, to thereby calculate medium density region average density values Mdx for the respective color components (Step S17).
[0077] In Fig. 12, the shadowed graph portions denote the cumulative density values which belong to the density section (DMk ≦ DM ≦ DMh) of the gray region cumulative density value tables. The total Σ(ΣDx)M of the cumulative density values ΣDx which are included in the shadowed graph portions is calculated. On the other hand, from the gray region average density frequency histograms, the total number ΣNpM of the pixels which exist in the density section is calculated. From Eq. 12 below, the average density value Mdx in the intermediate density region is found.
[B-2-3. Determining Correction Dummy Density Value]
[0078] Next, correction dummy density values (a dummy highlight density value and a dummy shadow density value for correction) are determined.
[0079] First, an average value Mda of the average density values Mdx for the respective color components in the intermediate density region is calculated
, and a difference MDx between the average value Mda and each average density value Mdx is then calculated (that is,
). The difference MDx shows the eccentricity amount from the average density value Mdx of each color component, and therefore, characterizes the amount of gray balance correction for each color in the intermediate density region.
[0080] Using these values as correction parameters and based on the highlight density value DHx and the dummy shadow density value DDSx which are calculated by the reference normalizing curve setting part 230 as reference density values, a correction dummy highlight density value CDHx and a correction dummy shadow density value CDSx for each color component are established from Eqs. 13 and 14 below (Step S18).
where KH and KS are proportional constants which are determined by experience depending on the type and the like of the original film 1.
[B-2-4. Setting Correction Function Gx]
[0081] Using the correction dummy highlight density value CDHx and the correction dummy shadow density value CDSx, a reference function correction part 248 corrects the nonlinear function Fx which indicates the characteristics of the reference normalizing curve C which is established in (B-1) above to thereby yield a correction function Gx which characterizes a correction normalizing curve Cr in which the gray balance in the intermediate density region is corrected (Step S19).
[0082] While the nonlinear function Fx is a function whose parameters are the density value D before normalization, the highlight density value DHx, the shadow density value DSx and the dummy shadow density value DDSx (that is, the nonlinear function Fx is expressed as Fx (D, DHx, DSx, DDSx)), the correction function Gx is set as Gx (D, DHx, DSx, CDHx, CDSx) since in place of the dummy shadow density value DDSx, the correction dummy shadow density value CDSx is used as a parameter, and the correction dummy highlight density value CDHx is also used as a parameter.
[0083] The correction function Gx must be established so that the correction normalizing curve Crx for each color component which will be determined by the correction function Gx satisfies the following conditions.
(a) Since the gray balance in the vicinity of the highlight density value and the shadow density value is already corrected using the reference normalizing curve C or needs not be corrected, on the normalization coordinates plane, the correction normalizing curve Crx passes through a first and a second points which are defined by these reference density values (First condition).
(b) In the intermediate density region, it is necessary to shift the reference normalizing curve C in a predetermined direction by the amount of correction of the gray balance. The amount of correction is dependent on MDx, i.e., the differences between the average density values Mdx and its average value Mda as described earlier, and therefore, the correction normalizing curve Crx runs close to or overlaps a straight line which passes through a third and a fourth points which are defined by the correction dummy shadow density value CDSx and the correction dummy highlight density value CDHx which are determined using the differences MDx as parameters (Second condition).
[0084] An example of the correction function Gx which satisfies these two conditions is expressed by Eq. 15 below.
where t is a proportional constant (coefficient) which is expressed by Eq. 16 below.
That is, the correction function Gx is expressed as a weighted mean of
and
. Curves expressed by the functions G1x and G2x have configurations as those which are shown in Fig. 13 as the curves C1x and C2x. The configurations of the curves shown in Fig. 13 are obtained when DHx < CDHx and CDSx < DSx in Eqs. 13 and 14.
[0085] The curve C1x is equal to the curve shown in Fig. 10 as it is amended by replacing the value DDSx with the value CDSx, and therefore, coincides with a straight line L1x in the intermediate and the low density regions. The curve C2x is equal to the curve shown in Fig. 10 as it is amended by further replacing the value DHx with the value CDHx, and therefore, coincides with a straight line L2x in the intermediate and the low density regions. The straight line L1x runs from a first point A1 which is defined as a two-dimensional coordinates point (DHx, px) and a third point A3 which is defined as a two-dimensional coordinates point (CDSx, qx). The straight line L2x runs from a fourth point A4 which is defined as a two-dimensional coordinates point (CDHx, px) and the third point A3.
[0086] On the other hand, from Eq. 16, the coefficient t is expressed as a secondary function of the density value D, and hence, a graph as that shown in Fig. 15, for example, is obtainable.
[0087] Since the correction function Gx is a weighted mean of the functions G1x and G2x as described earlier and 0 ≦ t ≦ 1 from the graph of Fig. 15, the correction normalizing curve Crx which is expressed by the correction function Gx passes between the curve C1x and the curve C2x.
[0088] The graph of Fig. 15 shows that t = 1 when
. Hence, from Eq. 15, it is found that
, thereby the correction normalizing curve Crx coinciding with the curve C1x and passing through the first point A1 (DHx, qx) which is defined by the highlight density value DHx. On the other hand, t = 0 when
, and therefore, from Eq. 15,
. Hence, the correction normalizing curve Crx passes through the second point A2 (DSx, px) which is defined by the shadow density value DSx in a similar manner.
[0089] Since the coefficient t is expressed as a secondary function of the density value D, as shown in Fig. 15, in the range DHx < Dx < DSx of the intermediate density region, a change in the coefficient t becomes larger as the density value D becomes smaller and smaller as the density value D becomes larger.
[0090] Eq. 15 also shows that the function G1x is multiplied by the coefficient t while the function G2x is multiplied by the coefficient (1 - t). It then follows that from a change ratio of the coefficient t, the correction normalizing curve Crx becomes close to the curve C2x from the first point A1, and as the density value D becomes larger, the correction normalizing curve Crx runs approximately along the curve C2x.
[0091] In the intermediate and the low density regions, the curve C2x coincides with the straight light L2x which runs from the fourth point A4 (CDHx, px) which is defined by the correction dummy highlight density value CDHx and the third point A3 (CDSx, qx) which is defined by the correction dummy shadow density value CDSx. Hence, the correction normalizing curve Crx becomes very close to or overlaps the straight light L2x (i.e., correction straight line) in the intermediate and the low density regions.
[0092] Fig. 14 is a graph which shows the correction normalizing curve Crx thus generated. As shown in Fig. 14, the correction normalizing curve Crx runs very close to or overlaps the straight line L2x which passes through the first point A1 defined by the highlight density value DHx and the second point A2 defined by the shadow density value DSx and through the third point A3 defined by the correction dummy shadow density value CDSx and the fourth point A4 defined by the correction dummy highlight density value CDHx in the intermediate and the low density regions. Thus, the correction normalizing curve Crx which is expressed by the correction function Gx satisfies both the first condition and the second condition.
[0093] It is to be noted that the correction function Gx defined by Eqs. 15 and 16 is merely an example. As far as satisfying the first condition and the second condition, any other appropriate function may be used as the correction function.
[0094] Once the correction function Gx is specified for each color component, numeric data which express the conversion characteristics of the correction function Gx, i.e., data in which the curve Crx of Fig. 14 is expressed by numerical values, are created and loaded in the LUT 202 of Fig. 1 for each color component.
[0095] Following this, as described earlier, based on the data loaded in the LUT 202, the density signals Dx of the respective pixels are normalized during scanning.
[0096] A comparison part 247 of the intermediate density region correction part 240 of Fig. 2 receives from the first cumulative processing part 232 and the second cumulative processing part 243 the total number of pixels of an inputted image-to-be-processed and the total number of pixels which are included in the gray region of the image-to-be-processed, respectively. If the ratio of the number of the pixels of the gray region to the total number of the pixels is equal to or smaller than a certain value (e.g., 10%), that is, if the original is occupied largely by primitive colors, since a demand for correction of the gray balance is small, the comparison part 247 automatically instructs the reference function correction part 248 not to perform correction of the gray balance.
〈C. Modifications〉
[0097]
(1) Correction of the reference normalizing curve C completes in the manner above.
However, in the normalizing curve generation part 210 according to this embodiment
shown in Fig. 2, since the reference normalizing curve C is created and thereafter
corrected if correction is necessary, it is also possible that information about an
already available normalizing curve is inputted from outside, corrected in the intermediate
density region correction part 240 and used.
If the already available normalizing curve does not have the dummy shadow density
value as the reference value, Eq. 14 needs be changed into Eq. 17 below.
In other words, where a proportional constant is KS1, it is possible to determine the correction dummy shadow density value CDSx using the shadow density value of the already available normalizing curve as the
reference value.
(2) In addition, it is possible that the same circuit is used commonly as the reference
normalizing curve setting part 230 and the intermediate density region correction
part 240 and that the function Gx which satisfies both the first condition and the second condition is directly generated
without creating the reference normalizing curve C.
Further, in the process color scanner 100 of Fig. 1, a portion or all of the normalizing
curve generation part 210 may be realized by the CPU 17 and the memory 18 of the information
processing apparatus 16.
(3) Although normalization is performed as a digital process using the LUT 202 in the embodiment above, normalization may be performed as an analog process using a function generator.
(4) It is desirable to determine the highlight density value DHx and the shadow density value DSx by a statistical method using a histogram and the like as is done in the embodiment above. However, if there are a highlight point and a shadow point in the original film 1, the color densities of the highlight point and the shadow point which are measured by an operator may be used as the highlight density value DHx and the shadow density value DSx.
(5) An encountered application of the present invention is not limited to a process color scanner. Rather, the present invention is applicable to any other image processing apparatuses which performs color separation and other operations on the image of an original.
(a) specifying gray pixels on said original color image;
(b) averaging said original color image signals Dx for respective color components in each gray pixel to obtain an average signal DM for each gray pixel;
(c) dividing a range of said average signal DM into a plurality of classes DMi;
(d) classifying said average signal DM representative of said gray pixels into said plurality of classes DMi;
(e) accumulating said original color signals Dx for each color component in each class DMi of said average signal DM to obtain accumulated density signal ΣDx for each color component;
(f) generating a sum signal Σ(ΣDx)M representative of a sum of said accumulated density signal ΣDx for each color component within a predetermined intermediate density section DMk ≦ DM ≦ DMh;
(g) counting number ΣNpn of pixels whose original color signals Dx are included in said intermediate density section DMk ≦ DM ≦ DMh;
(h) dividing said sum signal Σ(ΣDx)M for each color component by said number ΣNpn of pixels to obtain accumulated average signals Mdx for respective color components;
(i) averaging said accumulated average signals Mdx for respective color components to obtain a gray average signal Mda;
(j) generating difference signals
between said accumulated average signals Mdx and said gray average signal Mda;
(k) modifying a predetermined reference signal-conversion characteristic C as a function of said difference signals MDx = (Mdx - Mda) to obtain modified signal-conversion characteristics Crx for respective color components; and
(l) converting said original color image signals Dx by said modified signal-conversion characteristics Crx to obtain said converted color image signals DNx for respective color components.
the step (a) comprises the steps of:
(a-1) generating color-difference signals Dd representative of a difference among said original color image signals Dx for respective color components in each pixel; and
(a-2) comparing said color-difference signals Dd with a predetermined threshold signal in each pixel to specify said gray pixels.
the step (a) further comprises the steps of:
(a-3) determining a ratio of number of said gray pixels to total number of pixels on said original color image;
(a-4) comparing said ratio with a predetermined threshold ratio;
said step (k) comprising the step of:
(k- 1) modifying said predetermined reference signal-conversion characteristic C as
a function of said difference signals
only when said total number of said gray pixels is larger than said threshold pixel
number.
said step (k) further comprises the steps of:
(k-2) determining dummy highlight points CDHx and dummy shadow points CDSx of said original color image for respective color components as a function of said
difference signals
;
(k-3) determining linear signal-conversion characteristics L2x for respective color components defined on a coordinate plane and connecting; first
points (CDHx, px) defined by said dummy highlight densities CDHx and said first predetermined signal values px, and
second points (CDSx, qx) defined by said dummy shadow densities CDSx and said second predetermined signal values qx, and
(k-4) modifying said reference signal-conversion characteristic C so as to coincide with or approximate to said linear signal-conversion characteristics L2x in density sections between said dummy highlight densities CDHx and said dummy shadow densities CDSx to obtain said modified signal-conversion characteristics Crx for respective color components, respectively.
said step (f) comprises the step of:
(f-1) determining said intermediate density section DMk ≦ DM ≦ DMh as a function of statistical distributions of said average signals DM.
(a) determining a highlight density DHx and a shadow density DSx in a possible density range of said original color image signals Dx;
(b) determining dummy highlight densities CDHx and dummy shadow densities CDSx for respective color components in said possible density range of said original color image signals Dx;
(c) determining signal-conversion characteristics Crx for respective color components representative of signal-conversion curves defined on a coordinate plane satisfying the conditions that;
I) said signal-conversion curves pass through;
first points (DHx, px) defined by said highlight densities DHx and first predetermined signal values px, and
second points (DSx, qx) defined by said shadow densities DSx and second predetermined signal values qx, and
II) part of said signal-conversion curves defined for a predetermined intermediate
density section coincide with or approximate to lines connecting;
third points (CDHx, px) defined by said dummy highlight densities CDHx and said first predetermined signal values px, and
fourth points (CDSx, qx) defined by said dummy shadow densities CDSx and said second predetermined signal values qx, and
(d) converting said original color image signals Dx by said signal-conversion characteristics Crx to obtain said converted color image signals DNx for respective color components.
the step (b) comprises the step of:
(b-1) determining said dummy highlight densities CDHx and said dummy shadow densities CDSx as a function of a statistical distribution of gray pixels on said original color image.
(a) means for specifying gray pixels on said original color image;
(b) means for averaging said original color image signals Dx for respective color components in each gray pixel to obtain an average signal DM for each gray pixel;
(c) means for dividing a range of said average signal DM into a plurality of classes DMi;
(d) means for classifying said average signal DM representative of said gray pixels into said plurality of classes DMi;
(e) means for accumulating said original color signals Dx for each color component in each class DMi of said average signal DM to obtain accumulated density signal ΣDx for each color component;
(f) means for generating a sum signal Σ(ΣDx)M representative of a sum of said accumulated density signal ΣDx for each color component within a predetermined intermediate density section DMk ≦ DM ≦ DMh;
(g) means for counting number ΣNpn of pixels whose original color signals Dx are included in said intermediate density section DMk ≦ DM ≦ DMh;
(h) means for dividing said sum signal Σ(ΣDx)M for each color component by said number ΣNpn of pixels to obtain accumulated average signals Mdx for respective color components;
(i) means for averaging said accumulated average signals Mdx for respective color components to obtain a gray average signal Mda;
(j) means for generating difference signals
between said accumulated average signals Mdx and said gray average signal Mda;
(k) means for modifying a predetermined reference signal-conversion characteristic
C as a function of said difference signals
to obtain modified signal-conversion characteristics Crx for respective color components; and
(l) means for converting said original color image signals Dx by said modified signal-conversion characteristics Crx to obtain said converted color image signals DNx for respective color components.
the means (a) comprises:
(a-1) means for generating color-difference signals Dd representative of a difference among said original color image signals Dx for respective color components in each pixel; and
(a-2) means for comparing said color-difference signals Dd with a predetermined threshold signal in each pixel to specify said gray pixels.
the means (a) further comprises:
(a-3) means for determining a ratio of number of said gray pixels to total number of pixels on said original color image;
(a-4) means for comparing said ratio with a predetermined threshold ratio;
said means (k) comprising :
(k-1) means for modifying said predetermined reference signal-conversion characteristic
C as a function of said difference signals
only when said total number of said gray pixels is larger than said threshold pixel
number.
said means (k) further comprises:
(k-2) means for determining dummy highlight points CDHx and dummy shadow points CDSx of said original color image for respective color components as a function of said
difference signals
;
(k-3) means for determining linear signal-conversion characteristics L2x for respective color components defined on a coordinate plane and connecting;
first points (CDHx, px) defined by said dummy highlight densities CDHx and said first predetermined signal values px, and
second points (CDSx, qx) defined by said dummy shadow densities CDSx and said second predetermined signal values qx, and
(k-4) means for modifying said reference signal-conversion characteristic C so as to coincide with or approximate to said linear signal-conversion characteristics L2x in density sections between said dummy highlight densities CDHx and said dummy shadow densities CDSx to obtain said modified signal-conversion characteristics Crx for respective color components, respectively.
said means (f) comprises:
(f-1) means for determining said intermediate density section DMk ≦ DM ≦ DMh as a function of statistical distributions of said average signals DM.
(a) means for determining a highlight density DHx and a shadow density DSx in a possible density range of said original color image signals Dx;
(b) means for determining dummy highlight densities CDHx and dummy shadow densities CDSx for respective color components in said possible density range of said original color image signals Dx;
(c) means for determining signal-conversion characteristics Crx for respective color components representative of signal-conversion curves defined on a coordinate plane satisfying the conditions that;
I) said signal-conversion curves pass through;
first points (DHx, px) defined by said highlight densities DHx and first predetermined signal values px, and
second points (DSx, qx) defined by said shadow densities DSx and second predetermined signal values qx, and
II) part of said signal-conversion curves defined for a predetermined intermediate
density section coincide with or approximate to lines connecting;
third points (CDHx, px) defined by said dummy highlight densities CDHx and said first predetermined signal values px, and
fourth points (CDSx, qx) defined by said dummy shadow densities CDSx and said second predetermined signal values qx, and
(d) means for converting said original color image signals Dx by said signal-conversion characteristics Crx to obtain said converted color image signals DNx for respective color components.
the means (b) comprises:
(b-1) means for determining said dummy highlight densities CDHx and said dummy shadow densities CDSx as a function of a statistical distribution of gray pixels on said original color image.